Nvidia's Moat Crumbles: OpenAI's Jalapeño Chip and a 31% GPU Rental Crash Test Investor Faith

GPU rental prices plummet 31% as AI shifts to inference, OpenAI launches custom chip 'Jalapeño', and major customers develop proprietary hardware, but Nvidia's €119B backlog and AI factory pivot offer some relief.

The euphoria that once lifted Nvidia shares to dizzying heights is giving way to a harsher reality. GPU rental prices have plunged 31% in just three weeks, signaling that the frantic scramble for hardware is cooling. At the same time, the shift from training massive AI models to deploying them—known as inference—is reshaping the competitive landscape. Inference is expected to account for two-thirds of global AI compute by 2026, a transition that benefits custom chips over Nvidia's standard graphics processors. The stock ended the week at €168.80, according to one calculation, and at €168.94 in another—roughly 16% to 17% below its May record high. A 7% weekly loss reflected mounting anxiety among investors.

The most direct blow came on June 24, when OpenAI, the creator of ChatGPT, unveiled its first homegrown chip, named Jalapeño, developed in partnership with Broadcom. The chip is tailored for inference, the compute-intensive process of delivering AI results to users. Broadcom CEO Hock Tan claims such custom designs can cut costs by about 50% compared to off-the-shelf GPUs. What's more, OpenAI designed the chip in just nine months, using its own AI models to accelerate the process—a harbinger of a self-reinforcing cycle where AI speeds up the creation of AI hardware. OpenAI is not alone. Google, Amazon, Microsoft, and Meta are all developing proprietary chips, each a major Nvidia customer today and a potential rival tomorrow.

This customer defection compounds geopolitical headwinds. China once contributed a fifth of Nvidia's data-center revenue, but strict export controls have largely blocked that business. Local rivals like Huawei have filled the void. CEO Jensen Huang recently accompanied President Trump to China, highlighting Nvidia's delicate balancing act between seeking partnerships and complying with restrictions. Meanwhile, Chinese AI models such as GLM-5.2 from Z.ai are matching American benchmarks at a fraction of the cost, intensifying software competition and forcing Nvidia to rethink how its Blackwell architecture is deployed.

Yet the industrial side of the AI boom offers a counterweight. Nvidia is pivoting from pure data centers to specialized AI factories. Roche, the pharmaceutical giant, is using over 3,500 Blackwell chips to accelerate drug discovery, with the goal of moving molecules 25% faster through research. In Boden, Sweden, a massive data center is being built to lease up to 10,000 Nvidia chips to state and industrial clients. The company's order backlog stands at €119 billion, providing a cushion against market jitters. Nvidia's technology still powers more than 400 of the world's 500 fastest supercomputers, and nine in ten new systems in the latest ISC High Performance ranking rely on it. At the Hamburg conference, Nvidia positioned its new Vera-Rubin system as a scientific platform, and 35 new AI supercomputers are under construction across 23 European countries—a market that custom chips from cloud giants cannot easily serve.

Fundamentally, Nvidia remains on strong footing. Revenue hit a quarterly record of €81.6 billion, and the company has launched an €80 billion share buyback program alongside a quarterly dividend of $0.25 per share. Yet the stock's year-to-date gain stands at a modest 5% by one measure and 27% by another, reflecting investor uncertainty about the path ahead. The average analyst price target still hovers around €261, implying roughly 55% upside from Friday's close. These bullish projections rest on assumptions that Nvidia's software ecosystem will retain customers and that demand will remain large enough for both custom chips and GPUs.

Chart watchers see a technical test unfolding. The stock is trading barely 3% above its 200-day moving average of €163.66, a critical support level that, if breached, would darken the long-term outlook. The 100-day line at €168.66 is also being tested, and the relative strength index at 38.3 signals an almost oversold condition. Nvidia's weekly loss of 7% has shifted sentiment from optimism to caution.

What emerges is a company caught between two powerful currents. The first is a slowing GPU rental market and the rise of cheaper custom alternatives, exemplified by OpenAI's Jalapeño chip. The second is a deepening industrial pivot that could secure Nvidia's role as the backbone of scientific and enterprise computing. The moat that once seemed unbreachable now shows cracks, not collapse. But as customers start digging, investors are watching to see how deep those excavations go.